Get a Free Quote
aiShare Your Requirements
Technologies Involved:
PYTHON
Area Of Work: Generative AI
Project Description

Dudi Marketing, an Israel-based AI innovation firm specializing in autonomous systems approached Oodles to optimize and expand an early-stage Multi-Agent AI framework. The client aimed to define operational rules, integrate safeguards, and achieve seamless agent coordination to enhance their product’s market readiness and control. 
Scope Of Work

The client sought Oodles to address the challenge of maturing a loosely structured Multi-Agent system into a controlled, scalable AI ecosystem. The client sought Oodles for consulting and development support across agent communication design, safety protocols, rule-based governance, and modular system architecture. Key areas of work included agent logic orchestration, secure message handling, and ethical constraint modeling.
Our Solution

To meet the client's requirements, Oodles implemented a role-based Multi-Agent architecture layered with rule engines, communication protocols, and LLM safeguards.

Key features included:

  • Modular Agent Framework: Re-architected the existing system to assign roles like planner, executor, and auditor to individual agents, enhancing accountability and task-specific focus.
  • Rule Engine Development: Designed a flexible rule-based layer using LangChain and Python, enabling task-specific protocols, escalation conditions, and conflict prevention strategies.
  • Safety & Ethics Guardrails: Integrated LLMGuard and custom validators to enforce response boundaries. 
  • Real-Time Communication Protocols: Established WebSocket-based messaging with Redis and custom event queues.
  • Generative Memory & Context Handling: Leveraged OpenAI's GPT models with a VectorDB-backed memory to retain interaction history. 

Related Projects